AlgoDeal

[caementarius]

Quote from rolextrader:

Bottom Line: Only a fool would give an internet vendor his source code for a measly 10% even if the vendor was his next door neighbor. This sounds like a scheme to find a sucker who has profitable code and have it snaked away from a bunch of strangers offshore.

I've often thought there oughta be a way to solve the trust problem with technology but I suppose there is no magic bullet yet.

Looking up some background info on Algodeal, I think they have honorable backing from folks who would not want to be seen as sneaky:
http://www.corniche-group.com/page/logistics.html

 

[DeeDeeTwo]

Quote from rolextrader:

The deal still sucks, your explanation notwithstanding. Here's why: An investor usually pays a hedge fund 2/20 all the while the investor is at risk. The investor is not privy to details behind the trades, simply the global view of the system.

Then there is the issue of jurisdiction and enforcement in the absence of a binding agreement if you guys rip off the code. it's a financial nightmare to come after a French company.

Bottom Line: Only a fool would give an internet vendor his source code for a measly 10% even if the vendor was his next door neighbor. This sounds like a scheme to find a sucker who has profitable code and have it snaked away from a bunch of strangers offshore.

I'm trying to understand why random internet Algo Developers...
If there even is such a thing...
Would have any value to anyone.

Strategies have no value without expensive Execution infrastructure...
That means hardware, traders, engineers...
And ongoing maintenance daily and forever.

And your legal protection is zero...
As a general rule...
If you sign a multi-page contract with a firm 1000 times your size...
You are entirely at their mercy.

 

[caementarius]

Quote from elefevre:

Indeed. In fact, we expect that most people will first code and debug their strategy locally before submitting them (debugging remotely is not easy).
Open a strategy, then click the little arrow next to the Run button (green arrow at the top). Select "Run Strategy Local".

Oops, I've been effectively spamming a bunch of test cases to the remote run. (The one with a score in the 80s and 1.6 sharpe ratio is probably luck, don't run that).

It looks like I could convert minute data from other sources to run on locally as well. Can I expand the group of Stocks to include some custom ones even though they wouldn't be able to be backtested remotely?

This means the only drawdown to using Algodeal purely for personal use is it gives fewer statistics when run locally and doesn't have the grid optimization. From what I can tell, the grid stuff only helps when you are doing optimization.

 

[elefevre]

Quote from caementarius:
Can I expand the group of Stocks to include some custom ones even though they wouldn't be able to be backtested remotely?

Technically, yes, but it is a bit awkward. You'd need to extract the source for Stock (which we do provide in the download) and copy it in its usual package directory in your Eclipse project. My guess is that it should then take precedence over the one we provide, but you'll have to check that.

It is easier to take an existing stock and create a CSV file with the data you want, using the same format as the ones for the other Market Data files.

Quote from caementarius:
This means the only drawdown to using Algodeal purely for personal use is it gives fewer statistics when run locally and doesn't have the grid optimization. From what I can tell, the grid stuff only helps when you are doing optimization.

Well, that, and the fact that we provide a lot more data on the grid, and their quality is *much* higher.

We also have a nice (I think) feature on the grid that re-runs selected backtests with newer market data every day.

 

[Algo_Design_Kid]

It doesn't seem like it is a terrible idea from the POV of a "user". The front office documentation, lawyers, IT - I am sure all add up to quite a bit of cash.

When does "Algodeal" launch? I see on the website it says beta. Or are you currently running submitted strategies?

 

[elefevre]

Quote from Algo_Design_Kid:
When does "Algodeal" launch? I see on the website it says beta. Or are you currently running submitted strategies? [/B]

Our live system is still under development, although we do have a test strategy running on it. There is more work needed for a real production system. Maybe a few months?

Once we have a production-ready system, we will pitch strategies to our investor, on a case-by-case basis. Selection criteria are still being discussed, but they probably combine the basic strategies metrics, with the author's experience/track record, with our own understanding of what it does, with how easy it is to safeguard what it does, etc. We'll also need to develop a close relationship with the selected authors, as we expect to frequently discuss performances with them, as to decide whether to tune the strategy, etc.

As for removing the "beta" mention... I'm not sure. We might keep for a while, Google-style ;-)

 

[caementarius]

Quote from elefevre:

Our live system is still under development, although we do have a test strategy running on it. There is more work needed for a real production system. Maybe a few months?

Once we have a production-ready system, we will pitch strategies to our investor, on a case-by-case basis. Selection criteria are still being discussed, but they probably combine the basic strategies metrics, with the author's experience/track record, with our own understanding of what it does, with how easy it is to safeguard what it does, etc. We'll also need to develop a close relationship with the selected authors, as we expect to frequently discuss performances with them, as to decide whether to tune the strategy, etc.

elefevre -

I'll be glad to see when you all are making trades and running money. Until then, I really like the idea - especially if you are providing significant value.

Significant value would be, IMHO:

1) Ability to validate, as much as possible, a "good system", so that your capital connection is confident
2) Ability to get size for a "good system" -- i.e., if you start a great strategy with $100k, then $500k 6 months later, more 6 months later, I'm not sure that is so great. While bootstrapping and connecting with capital might be hard, it's not impossible. In fact, I'm pretty sure if I had the system I could find the money on my own. $1M to $5M would be more reasonable.

Also, add ETFs when you can. How hard can it be? I think ETFs are used a fair amount by system developers.

 

[Algo_Design_Kid]

Visited the site a while back and signed up. Only been back a few times - the website , updates , features and activity on the website seems dead?

I also see in the support page unanswered questions that have been posted to administrators for weeks now.

It seems as though they took all the ideas submitted and just decided to pack up shop. Who didn't see that coming?

Thank god I didn't find out about this any earlier and actually submit any real work. Feel bad for whomever they conned in this process.

 

[elefevre]

Quote from Algo_Design_Kid:
Visited the site a while back and signed up. Only been back a few times - the website , updates , features and activity on the website seems dead?

We had a release two weeks ago.

I also see in the support page unanswered questions that have been posted to administrators for weeks now.

It seems as though they took all the ideas submitted and just decided to pack up shop. Who didn't see that coming?

Well, that's jumping to conclusions, don't you think?
First of all, we have not closed shop.
Second, concerning the threads on the support site, in many cases, the questions were answered, but were reopened automatically as someone added more information. I have now closed most of them again.

Anyway, in the past few months we have been in contact with a limited number of quants (not necessarily through the website). They have provided us with strategies that have much higher quality than what we find on our public backtest platform. So far, only a very small number of users of the platform have submitted strategies that are worth investigating more. We have started discussing with them too.

Our website will reflect this change of strategy in the coming weeks.

If others feel that we should have been discussing with them as well, feel free to get in touch (our contact details are on the website).

 

[caementarius]

Quote from elefevre:

Anyway, in the past few months we have been in contact with a limited number of quants (not necessarily through the website). They have provided us with strategies that have much higher quality than what we find on our public backtest platform. So far, only a very small number of users of the platform have submitted strategies that are worth investigating more. We have started discussing with them too.

Our website will reflect this change of strategy in the coming weeks.

If others feel that we should have been discussing with them as well, feel free to get in touch (our contact details are on the website).

I suspect that you might see more work from the crowd once you start trading strategies and cutting checks. I'd like to spend more time with it, and I probably will, but I'm pretty occupied.

I think most folks who can work up a good strategy can work out the technical details of backtesting and trading it - which leaves the main advantage of algodeal being connections with capital.

However, if you wait a year before scaling up, that's roughly the amount of time (depending on the strategy) that other capital would want to see in terms of historical performance. So, it's a little less exciting to put a lot of sweat into something for the chance of 10% on whatever is made with $100k.

Unless you are targeting ostensibly smart quants who are totally off the grid, and who can't come up with $10k -- which I think would be relatively rare since I'm sure smart poor people who can put the work in would not be poor for very long.

I believe in the idea, and put together some unoriginal examples that score in the 80s, but at some point you will probably need to spend more time/effort building relationships.